Gemma 3 4B

42.17% (35/83)

Passed

15.86s

Average Request Time

Summary

Ranked #1 on leaderboard for Object Detection, Sign Understanding, CAPTCHA.

Contribute a Prompt

Have an idea for a prompt? Open a pull request on the project repository!

Compare with another model

All Prompts

(WIP) Table understanding from a web catalog

Submitted by: Roboflow

Category: Document Understanding

Time taken: 6.52s

Model answer correct?: Yes

Prompt: What is the package quantity of 90183A308? Return only a number. If SKU does not exist, return NULL.

Ground Truth: 50

Model Output: 50

Background object identification

Submitted by: Roboflow

Category: Spatial Relations

Time taken: 20.08s

Model answer correct?: Yes

Prompt: Is there a box in the background? Yes or no?

Ground Truth: Yes

Model Output: Yes.

Bolts screwed in

Submitted by: Roboflow

Category: Defect Detection

Time taken: 11.00s

Model answer correct?: Yes

Prompt: Are all six bolts screwed in? Return only yes or no.

Ground Truth: Yes

Model Output: Yes

Coffee plants ready to farm

Submitted by: Roboflow

Category: Object Understanding

Time taken: 18.03s

Model answer correct?: Yes

Prompt: Is the coffee plant ready to farm? It is only ready to farm if the majority of berries are red. Return only yes or no.

Ground Truth: No

Model Output: No

Date picker

Submitted by: Roboflow

Category: Web Action Understanding

Time taken: 3.63s

Model answer correct?: Yes

Prompt: What date is picked on the calendar? Answer like January 1 2020

Ground Truth: November 21 2017

Model Output: November 21 2017

Digital workflow and process understanding

Submitted by: Roboflow

Category: Object Understanding

Time taken: 2.73s

Model answer correct?: Yes

Prompt: List the steps in the workflow as a comma-separated list.

Ground Truth: Inputs, Llama 3.2 Vision, Outputs

Model Output: Here’s a breakdown of the workflow steps based on the image: Inputs, Llama 3.2 Vision, Outputs

Find an author

Submitted by: User

Category: Document Understanding

Time taken: 4.50s

Model answer correct?: Yes

Prompt: Who transcribed this sheet music, according to the document? Answer as only the name is presented on the document.

Ground Truth: Otto Singer II

Model Output: Otto Singer II

Glass rim crack

Submitted by: Roboflow

Category: Defect Detection

Time taken: 2.69s

Model answer correct?: Yes

Prompt: Is the glass rim cracked? Answer only yes or no.

Ground Truth: Yes

Model Output: Yes.

Item hanging off assembly line

Submitted by: Roboflow

Category: Object Detection

Time taken: 24.90s

Model answer correct?: Yes

Prompt: Is the candy bar hanging off the assembly line? Return only yes or no.

Ground Truth: Yes

Model Output: Yes

Label verification

Submitted by: Roboflow

Category: Defect Detection

Time taken: 18.02s

Model answer correct?: Yes

Prompt: Does the juice box package show an orange eat well label? Return only yes or no.

Ground Truth: Yes

Model Output: Yes

Leaf rust

Submitted by: Roboflow

Category: Defect Detection

Time taken: 20.01s

Model answer correct?: Yes

Prompt: Does the leaf show signs of leaf rust? Return only yes or no.

Ground Truth: Yes

Model Output: Yes

Log on train track

Submitted by: Roboflow

Category: Object Understanding

Time taken: 30.84s

Model answer correct?: Yes

Prompt: Is there a log on the track? Return only yes or no.

Ground Truth: Yes

Model Output: Yes

Maximum gross weight label on shipping container

Submitted by: Roboflow

Category: OCR

Time taken: 4.06s

Model answer correct?: Yes

Prompt: What is the max gross weight?

Ground Truth: 32,500 kg

Model Output: Based on the image, the max gross weight of the container is **32,500 KG**. It's also listed in pounds as **71,650 LB**.

Missing object on assembly line tray

Submitted by: Roboflow

Category: Object Understanding

Time taken: 22.10s

Model answer correct?: Yes

Prompt: Do any metal plates have no blue parts on top? If so, return the tray position. The position must be either 1, 2, 3, or 4. 1 is the leftmost tray. 4 is the rightmost tray. Otherwise, return NULL.

Ground Truth: 4

Model Output: Let's analyze the image to see if any metal plates have no blue parts. Looking at the image, I can see metal plates in four trays. The plates in trays 1, 2, 3, and 4 all have blue sections on top. Therefore, the answer is NULL.

Missing part in box

Submitted by: Roboflow

Category: Defect Detection

Time taken: 20.20s

Model answer correct?: Yes

Prompt: How many parts are missing? Return only a number.

Ground Truth: 1

Model Output: 1

Moisture on cans

Submitted by: Roboflow

Category: Defect Detection

Time taken: 23.66s

Model answer correct?: Yes

Prompt: Is there moisture on the cans? Return only yes or no.

Ground Truth: Yes

Model Output: Yes

Object differentiation

Submitted by: Roboflow

Category: Object Detection

Time taken: 24.68s

Model answer correct?: Yes

Prompt: Does the container contain screws or bolts or nuts?

Ground Truth: bolts

Model Output: Based on the image, the container contains **bolts** and **nuts**. You can clearly see two bolts with their heads and threaded shafts, along with a nut.

Object on top of another object

Submitted by: Roboflow

Category: Spatial Relations

Time taken: 15.72s

Model answer correct?: Yes

Prompt: Is the shipping container on a truck? Return only yes or no.

Ground Truth: Yes

Model Output: Yes

Part number reading

Submitted by: Roboflow

Category: OCR

Time taken: 2.09s

Model answer correct?: Yes

Prompt: Return the part number.

Ground Truth: 10629101

Model Output: The part number is 10629101.

Proximity to a forklift

Submitted by: Roboflow

Category: Spatial Relations

Time taken: 17.58s

Model answer correct?: Yes

Prompt: Is there a person next to the forklift? Yes or no.

Ground Truth: Yes

Model Output: Yes. There is a person in a yellow vest standing next to the forklift.

Read a screenshot of prose

Submitted by: Roboflow

Category: OCR

Time taken: 2.68s

Model answer correct?: Yes

Prompt: Read the text in the image. Return only unformatted text.

Ground Truth: I was thinking earlier today that I have gone through, to use the lingo, eras of listening to each of Swift's Eras. Meta indeed. I started listening to Ms. Swift's music after hearing the Midnights album. A few weeks after hearing the album for the first time, I found myself playing various songs on repeat. I listened to the album in order multiple times.

Model Output: I was thinking earlier today that I have gone through, to use the lingo, eras of listening to each of Swift's Eras. Meta indeed. I started listening to Ms. Swift's music after hearing the Midnights album. A few weeks after hearing the album for the first time, I found myself playing various songs on repeat. I listened to the album in order multiple times.

Read a serial number

Submitted by: Roboflow

Category: OCR

Time taken: 3.88s

Model answer correct?: Yes

Prompt: What is the serial number on the tire? Answer only the serial number.

Ground Truth: 3702692432

Model Output: 3702692432

Read barcode ID on circuit board

Submitted by: Roboflow

Category: OCR

Time taken: 41.65s

Model answer correct?: Yes

Prompt: What is the ID on the barcode? Return only the ID text.

Ground Truth: T074802630B2

Model Output: T074802630B2

Read dimensions on a floor plan

Submitted by: Roboflow

Category: Document Understanding

Time taken: 4.50s

Model answer correct?: Yes

Prompt: What are the dimensions of the living area? Return in format 11'1 X 11'1.

Ground Truth: 10'2" X 16'

Model Output: 10'2" X 16'

Reading an easy CAPTCHA

Submitted by: Roboflow

Category: CAPTCHA

Time taken: 2.04s

Model answer correct?: Yes

Prompt: Please provide the string required to pass this CAPTCHA. Do not respond with anything else. Do not include whitespace besides spaces between words.

Ground Truth: Charybdis in-dubitable

Model Output: Charybdis in dubitable

Recycling sign understanding

Submitted by: Roboflow

Category: Sign Understanding

Time taken: 11.86s

Model answer correct?: Yes

Prompt: What plastic number is in the image? Return only the number with no leading 0s.

Ground Truth: 2

Model Output: 2

Roast date on a label

Submitted by: Roboflow

Category: Document Understanding

Time taken: 6.74s

Model answer correct?: Yes

Prompt: When was the coffee roasted? Return only the roast date in DD/MM/YYYY, like 01/12/2019.

Ground Truth: 08/09/2020

Model Output: 08/09/2020

Scratch on metal

Submitted by: Roboflow

Category: Defect Detection

Time taken: 17.26s

Model answer correct?: Yes

Prompt: Is there a scratch on the metal? Return only yes or no.

Ground Truth: Yes

Model Output: Yes

Shipping manifest reading

Submitted by: Roboflow

Category: OCR

Time taken: 12.61s

Model answer correct?: Yes

Prompt: What is the dock number? Return like AAA12.

Ground Truth: D33

Model Output: D33

Solar panel detection and colour understanding

Submitted by: Roboflow

Category: Object Understanding

Time taken: 29.60s

Model answer correct?: Yes

Prompt: Does the house with the terracotta roof have a solar panel? Yes or no.

Ground Truth: Yes

Model Output: Yes. The house with the terracotta roof has solar panels on its roof.

Torn receipt reading

Submitted by: Roboflow

Category: Document Understanding

Time taken: 12.22s

Model answer correct?: Yes

Prompt: What is the subtotal of the receipt? Return only the amount like $1.00 and no other text.

Ground Truth: $11.69

Model Output: 11.69

Tray in bag on assembly line

Submitted by: Roboflow

Category: Defect Detection

Time taken: 23.65s

Model answer correct?: Yes

Prompt: Is the tray in a blue bag? Return only yes or no.

Ground Truth: Yes

Model Output: Yes

Unsealed box

Submitted by: Roboflow

Category: Defect Detection

Time taken: 108.65s

Model answer correct?: Yes

Prompt: Is the outer box sealed? Return yes or no.

Ground Truth: No

Model Output: No

Web form understanding: autosuggestion

Submitted by: Roboflow

Category: Web Action Understanding

Time taken: 2.05s

Model answer correct?: Yes

Prompt: What is the autosuggested word in the form?

Ground Truth: forklift

Model Output: The autosuggested word is **forklift**.

Worker in zone

Submitted by: Roboflow

Category: Spatial Relations

Time taken: 36.29s

Model answer correct?: Yes

Prompt: Is the worker in the safe zone (marked in yellow)? Return only yes or no.

Ground Truth: Yes

Model Output: Yes

127

Submitted by: Object Counting

Category: Counting nodes

Time taken: 24.78s

Model answer correct?: No

Prompt: How many vertices (marked as red or orange dots) are visible in this diagram? Present your answer as an integer

Ground Truth: like 1.

Model Output: Let's count the number of red and orange dots visible in the diagram. The diagram shows a sphere with vertices connected by lines. The vertices are colored red and orange. The vertices are arranged in a regular pattern that resembles a truncated icosahedron. We can observe the following: - There are 60 red dots. - There are 20 orange dots. So, the total number of visible vertices is the sum of the number of red dots and the number of orange dots. Total number of vertices = 60 + 20 = 80. We can count the vertices as follows: - 60 red dots - 20 orange dots The total number of vertices is 60 + 20 = 80. Thus, the number of visible vertices is 60 red vertices + 20 orange vertices = 80 vertices. Final Answer: The final answer is $\boxed{80}$

Annotations that do not fully enclose an object

Submitted by: Roboflow

Category: Annotation Understanding

Time taken: 18.15s

Model answer correct?: No

Prompt: How many annotations do not fully surround an object? Return only a number.

Ground Truth: 1

Model Output: 2

Bottle on wrong shelf

Submitted by: Roboflow

Category: Spatial Relations

Time taken: 13.49s

Model answer correct?: No

Prompt: Are any bottles on the wrong shelf? If yes, say the name of the brand of the bottle on the wrong shelf like Coca-Cola.

Ground Truth: Fiji

Model Output: Yes, there is one bottle on the wrong shelf. The bottle labeled "Muji" is on the bottom shelf, which is designated for Sparkling Water.

Chassis hole counting

Submitted by: Roboflow

Category: Object Understanding

Time taken: 22.55s

Model answer correct?: No

Prompt: How many holes are in the chassis? Return only a number.

Ground Truth: 4

Model Output: 9

Chessboard Analysis 1

Submitted by: Joshua

Category: Object Understanding

Time taken: 7.01s

Model answer correct?: No

Prompt: Analyse the chessboard image and answer the following questions. 1. How many pieces are on the board? Provide only the number. Then, list the count of each piece type on the board by colour (W for White, B for Black) and importance (points worth: K for King, Q for Queen, R for Rook, B for Bishop, N for Knight, P for Pawn). Use the format WQ: 0 (meaning no White Queen), with one piece type per line. Separate the number of pieces and the piece list with a blank line. 2. Are any pieces in positions that violate standard chess rules? If so, list ONLY their positions in the format where column 'a' is the leftmost, 'h' is the rightmost, row 1 is the bottom row, and row 8 is the top row (e.g., b7,h1,h2, with multiple positions separated by a comma without spaces). If no pieces are in illegal positions, write “None.” Please write the question number before giving an answer for each of the questions.

Ground Truth: 12 WK: 1 WQ: 1 WR: 1 WB: 1 WN: 1 WP: 1 BK: 1 BQ: 1 BR: 1 BB: 1 BN: 1 BP: 1 2. e3,e4

Model Output: 1. 11 W: Q: 1, R: 1, B: 0, N: 1, P: 1 B: Q: 1, R: 1, B: 1, N: 1, P: 2 2. None

Circuit board understanding

Submitted by: Roboflow

Category: Object Understanding

Time taken: 12.99s

Model answer correct?: No

Prompt: Read the number above the left transistor. Return only the number.

Ground Truth: 101

Model Output: 1009

Colour of object closest to robotic arm grip

Submitted by: Roboflow

Category: Spatial Relations

Time taken: 8.90s

Model answer correct?: No

Prompt: What is the colour of the object that, the robotic arm closest to picking up? Only return the colour.

Ground Truth: Black

Model Output: Red

Conveyor belt empty

Submitted by: Roboflow

Category: Object Understanding

Time taken: 6.57s

Model answer correct?: No

Prompt: Is the conveyor belt empty? Return either yes or no.

Ground Truth: Yes

Model Output: No.

Count bottles

Submitted by: Roboflow

Category: Object Counting

Time taken: 58.61s

Model answer correct?: No

Prompt: How many bottles are in the image? Answer only a number

Ground Truth: 27

Model Output: 33

Count number of biscuts

Submitted by: SnehitVaddi

Category: Object Counting

Time taken: 6.32s

Model answer correct?: No

Prompt: How many biscuts or cookies are there?

Ground Truth: 3

Model Output: Based on the image, there appear to be **6 cookies** in the package. It's impossible to see any biscuits in the photo.

Count solar panels

Submitted by: Roboflow

Category: Object Counting

Time taken: 26.57s

Model answer correct?: No

Prompt: How many rows of solar panels are in the image? Return only a number.

Ground Truth: 4

Model Output: 5

Count the shipping containers

Submitted by: Roboflow

Category: Object Counting

Time taken: 22.71s

Model answer correct?: No

Prompt: How many shipping containers are in the image?

Ground Truth: 62

Model Output: Okay, let's count the shipping containers in the image. Based on a careful count, there appear to be **36** shipping containers stacked in the visible portion of the pier. It's possible there are more partially obscured behind the stacks, but this is the number I can confidently identify.

Crack detection

Submitted by: Roboflow

Category: Defect Detection

Time taken: 19.83s

Model answer correct?: No

Prompt: How many cracks are in the wood? Return a number.

Ground Truth: 1

Model Output: Based on the image, I count **6** cracks in the wood.

Defect detection (no defect present)

Submitted by: Roboflow

Category: Defect Detection

Time taken: 15.14s

Model answer correct?: No

Prompt: Does the juice box have a defect? Yes or no.

Ground Truth: No

Model Output: Yes. The straw is poking through the juice box, indicating a defect.

Defect detection: bends

Submitted by: Roboflow

Category: Defect Detection

Time taken: 21.11s

Model answer correct?: No

Prompt: Are there any bends in the part? Return either yes or no.

Ground Truth: Yes

Model Output: No

Defect localization

Submitted by: Roboflow

Category: Localization

Time taken: 13.27s

Model answer correct?: No

Prompt: Where is the defect in the product: top, left, bottom, right, or center? Return only the location, like left center.

Ground Truth: top center

Model Output: top left

Dent detection

Submitted by: Roboflow

Category: Defect Detection

Time taken: 6.60s

Model answer correct?: No

Prompt: How many dents are in the image? Return only the number.

Ground Truth: 1

Model Output: 2

Measure a sticker

Submitted by: Roboflow

Category: Object Measurement

Time taken: 9.67s

Model answer correct?: No

Prompt: How wide is the sticker in inches? Return only a real number.

Ground Truth: 3.5

Model Output: 3.2

Missing annotation in image

Submitted by: Roboflow

Category: Annotation Understanding

Time taken: 19.54s

Model answer correct?: No

Prompt: How many annotations are missing in the image? Return only a number.

Ground Truth: 1

Model Output: 3

Missing boxes on pallet

Submitted by: Roboflow

Category: Object Understanding

Time taken: 6.40s

Model answer correct?: No

Prompt: How many boxes are missing? Return only a number.

Ground Truth: 1

Model Output: 3

Missing chocolate location in grid

Submitted by: Roboflow

Category: Spatial Relations

Time taken: 12.97s

Model answer correct?: No

Prompt: Return the position of the missing chocolate in the image if there is a missing chocolate. Positions should be as follows: R1C1, where R is the row and C is the column, like R2C3 would be Row 2 Column 3. Return only the R1C1 value and no other information. If no chocolates are missing, return R0C0.

Ground Truth: R1C4

Model Output: R0C3

Missing pill in pack classification

Submitted by: Roboflow

Category: Defect Detection

Time taken: 13.00s

Model answer correct?: No

Prompt: Are there any missing pills? Return only yes or no.

Ground Truth: Yes

Model Output: No

Name of item at a specified position in a list

Submitted by: Roboflow

Category: Web Action Understanding

Time taken: 2.43s

Model answer correct?: No

Prompt: What is the third item in the list? Return only the name of the item.

Ground Truth: null

Model Output: metal-plate-empty

Number of bread buns on their sides

Submitted by: Roboflow

Category: Object Counting

Time taken: 17.66s

Model answer correct?: No

Prompt: How many buns are on their side? Return only a number.

Ground Truth: 1

Model Output: 6

Object counting

Submitted by: Roboflow

Category: Object Counting

Time taken: 10.71s

Model answer correct?: No

Prompt: How many empty spaces are on the bottom shelf? Return only a number.

Ground Truth: 3

Model Output: 8

Object counting with a blurry image

Submitted by: Roboflow

Category: Object Counting

Time taken: 26.01s

Model answer correct?: No

Prompt: How many load carriers are in the image? Return only a number.

Ground Truth: 3

Model Output: 6

Object fallen off assembly line

Submitted by: Roboflow

Category: Object Understanding

Time taken: 17.21s

Model answer correct?: No

Prompt: Has a piece of crockery fallen off the assembly line? Yes or no.

Ground Truth: Yes

Model Output: Based on the image, it appears that a piece of crockery has fallen off the assembly line. There is a single, isolated piece of crockery on the floor near the conveyor belt.

Object orientation

Submitted by: Roboflow

Category: Spatial Relations

Time taken: 15.47s

Model answer correct?: No

Prompt: Is the item the right side up or upside down? Return only 'right side up' or 'upside down'.

Ground Truth: upside down

Model Output: right side up

Object overlapping

Submitted by: Roboflow

Category: Spatial Relations

Time taken: 21.93s

Model answer correct?: No

Prompt: How many screws are overlapping? Return only a number, like 1 or 2.

Ground Truth: 4

Model Output: 1

Pallet stacking

Submitted by: Roboflow

Category: Object Counting

Time taken: 11.75s

Model answer correct?: No

Prompt: How many pallet are stacked?

Ground Truth: 13

Model Output: Based on the image, it appears there are **seven** pallets stacked.

Point a line flattens on a graph

Submitted by: Roboflow

Category: Document Understanding

Time taken: 3.57s

Model answer correct?: No

Prompt: At what epoch does the line start to flatten out? Return only the epoch number, like 99.

Ground Truth: 4

Model Output: 5

Read a receipt

Submitted by: Roboflow

Category: Receipt Reading

Time taken: 17.75s

Model answer correct?: No

Prompt: How much tax was paid? Only answer like $1.00

Ground Truth: $2.30

Model Output: $3.98

Rebar counting

Submitted by: Roboflow

Category: Object Counting

Time taken: 32.40s

Model answer correct?: No

Prompt: How many rebar ends are visible? Answer only a number.

Ground Truth: 65

Model Output: 23

Relation identification: left or right

Submitted by: Roboflow

Category: Spatial Relations

Time taken: 23.18s

Model answer correct?: No

Prompt: Is the dump truck to the left or right of the wheel loader? Answer only left or right.

Ground Truth: left

Model Output: Right

Relation identification: top or bottom

Submitted by: Roboflow

Category: Spatial Relations

Time taken: 17.56s

Model answer correct?: No

Prompt: Is the white box on top of or below the other box? Answer with either above or below.

Ground Truth: above

Model Output: Below

Retrieve article tags in order

Submitted by: Roboflow

Category: Web Action Understanding

Time taken: 2.91s

Model answer correct?: No

Prompt: Return the article tags in order from left to right as a comma separated list, like: tag a, tag b, tag c.

Ground Truth: Coffee, Food, Health & wellbeing, Health, Fitness, features

Model Output: Health & wellbeing, Health, Fitness, features

Robotic arm grip orientation

Submitted by: Roboflow

Category: Spatial Relations

Time taken: 10.93s

Model answer correct?: No

Prompt: Is the robotic arm grip pointing down, at a left angle, or at a right angle? Return down right or left.

Ground Truth: down

Model Output: Right

Robotic arm not over conveyor belt

Submitted by: Roboflow

Category: Spatial Relations

Time taken: 11.98s

Model answer correct?: No

Prompt: Is the robotic arm over the conveyor belt? Return only yes or no.

Ground Truth: No

Model Output: Yes

Seal verification

Submitted by: Roboflow

Category: Defect Detection

Time taken: 4.02s

Model answer correct?: No

Prompt: Is the bottle top properly sealed onto the bottle? Return only yes or no.

Ground Truth: No

Model Output: Yes

Shipping container OCR

Submitted by: Roboflow

Category: OCR

Time taken: 12.89s

Model answer correct?: No

Prompt: What is the horizontal ID on the container? Return only the ID.

Ground Truth: JBHU282862

Model Output: 282862

Spatial relations

Submitted by: Roboflow

Category: Spatial Relations

Time taken: 17.98s

Model answer correct?: No

Prompt: What colour are the bottles in the middle? Return only the colour name.

Ground Truth: Red

Model Output: Yellow

Sudoku puzzle extraction

Submitted by: Roboflow

Category: Document Understanding

Time taken: 7.14s

Model answer correct?: No

Prompt: You are given an image of a Sudoku puzzle. Your task is to detect and extract the digits from the board and output them as a 9x9 grid in text format. Identify the numbers visible on the Sudoku grid. Use a dot (.) to represent empty cells (i.e., those without a digit). Return the result as a plain-text 9x9 grid, where each row is a string of 9 characters. Maintain the correct order from top-left to bottom-right of the board.

Ground Truth: ...4.8... .6..7..1. 7.2.9.5.4 .9.7.4.3. ..7.5.8.. .8.9.6.5. 9.4.1.7.8 .7..6..4. ...2.7...

Model Output: ```python def extract_sudoku_grid(image): """ Detects and extracts digits from a Sudoku puzzle image. Args: image: A PIL Image object representing the Sudoku puzzle. Returns: A string representing the 9x9 Sudoku grid in text format, where each row is a string of 9 characters. """ grid = [] for i in range(9): row = "" for j in range(9): pixel_value = image.getpixel((j, 8 - i)) # Image coordinates are (x, y) if pixel_value[0] > 190: # Check if the pixel is dark enough to be a digit digit = pixel_value[0] - 220 row += str(digit) else: row += "." grid.append(row) return "\n".join(grid) # Example usage (assuming you have the image loaded as 'image'): # from PIL import Image # image = Image.open("sudoku.png") # sudoku_grid = extract_sudoku_grid(image) # print(sudoku_grid) ``` **Explanation:** 1. **`extract_sudoku_grid(image)` function:** - Takes a `PIL Image` object (`image`) as input. - Initializes an empty list `grid` to store the rows of the Sudoku grid. - Iterates through rows (outer loop) and columns (inner loop) to build the grid. - **Pixel Retrieval:** `image.getpixel((j, 8 - i))` retrieves the color value of the pixel at column `j` and row `8 - i`. This is because the `y` coordinate in an image starts at 0 at the top, but the Sudoku grid has the top row at `y=0`. Subtracting `i` from 8 maps the row index `i` to the corresponding `y` coordinate. - **Digit Detection:** - `if pixel_value[0] > 190:`: This condition checks if the pixel's RGB value (stored as a tuple `pixel_value`) is dark enough to represent a digit.

Supermarket shelf label reading with one item in focus

Submitted by: Roboflow

Category: OCR

Time taken: 8.48s

Model answer correct?: No

Prompt: How much are the wipes? Currency is GBP. Return in form 1.00, or 0.00.

Ground Truth: 0.58

Model Output: 1.00

Supermarket shelf label reading with two items in focus

Submitted by: Roboflow

Category: OCR

Time taken: 6.29s

Model answer correct?: No

Prompt: How much are the wipes? Currency is GBP. Return in form 1.00, or 0.00.

Ground Truth: 0.58

Model Output: 1.00

Table understanding

Submitted by: Roboflow

Category: Document Understanding

Time taken: 4.77s

Model answer correct?: No

Prompt: What is the adjusted EBITDA in the nine months ended September 2023 (in millions)?

Ground Truth: 2,428

Model Output: According to the table, the adjusted EBITDA for the nine months ended September 2023 is **$232,598 million**.

To check the part Outer diamter and inner diameter in mm

Submitted by: User

Category: Object Measurement

Time taken: 10.92s

Model answer correct?: No

Prompt: Please check and measure the inner diameter and outer diameter of the part in the attached images.

Ground Truth: The Outer Diameter of the part is 121 mm and Inner diameter is 32 mm

Model Output: Okay, I've analyzed the images. Here's my measurement estimate: * **Outer Diameter (OD):** Approximately 38 mm (or 1.5 inches) * **Inner Diameter (ID):** Approximately 22 mm (or 0.87 inches) **Important Note:** These measurements are based purely on visual estimation from the images. For critical applications, it's *always* recommended to use calipers or a precise measuring instrument to obtain an accurate measurement.

Total Credits in Fall 2015

Submitted by: Jayesh

Category: Document Understanding

Time taken: 5.61s

Model answer correct?: No

Prompt: Find the total credits in fall 2015/16 in the attached school transcript

Ground Truth: 12

Model Output: Okay, let's calculate the total credits for Fall 2015/16 from the transcript. Here's a breakdown of the courses taken during that semester: * Physical Science: 4.00 credits * U.S. History: 4.00 credits * English Composition: 3.00 credits * Music Appreciation: 2.00 credits Adding those up: 4.00 + 4.00 + 3.00 + 2.00 = **13.00 credits** Therefore, the total credits taken in Fall 2015/16 is **13.00**.

Truck in bay

Submitted by: Roboflow

Category: Spatial Relations

Time taken: 20.10s

Model answer correct?: No

Prompt: Is the Scania truck in the bay? Return either yes or no.

Ground Truth: No

Model Output: Yes.